Lift restriction of only one weak connection per cycle
The following discussion from !88 (merged) should be addressed:
-
@eikeschulte started a discussion: I’m wondering whether this is necessary at all. There are certainly connection setups for which this can’t be achieved. (For example, three simulators with connections going back and forth between all of them.) Lifting this rule might introduce some non-determinism but this might be preferable to outlawing it at all.
As far as I can see, we only need that the dataflow graph with all time-shifted and weak connections removed can be topologically sorted (i.e. is a DAG).
This might be a topic for the developer meeting.